Method of using structural models for optical recognition

A physical storage device having stored computer-executable instructions that, when executed by a processor of a computing system that is in communication with the physical storage device, cause the computing system to perform a method comprising:

• obtaining at least one structural description and at least one vector representation for each object in an image, wherein each structural description includes a spatial-metrical relationship for each object relative to at least another object;
• creating a vector structural model for each object;
• creating a classifier, wherein the classifier is capable of comparing variants for each object against the object's vector structural model;
• comparing, by the classifier, each object against its respective vector structural model;
• and
• identifying a variant that best matches the object's vector structural model.

A computer system comprising:

• at least one processor;
• a sensor for receiving an image;
• system memory;
• and
• stored computer-executable instructions which, when executed by the at least one processor, implement a method for identifying objects in the image, the method comprising:
• receiving the image by the sensor;
• isolating an object from the image;
• preliminarily assigning a plurality of geometric structural elements and creating at least one structural model of each different object of a plurality of objects that could be found in the image, wherein a structural model describes an object as one consisting of at least one geometric structural element;
• providing at least one hypothesis about the isolated object;
• testing the at least one hypothesis by matching the isolated object with structural models of the at least one generated hypothesis, wherein structural elements described in a structural model are purposefully searched in the image;
• and
• presenting an indication of a result of the testing of at least one hypothesis.